Is your feature request related to a problem? Please describe.
I just discovered the following repository : https://github.com/vicgalle/stable-diffusion-aesthetic-gradients
It implements "aestetic embeddings" that are, if I understood well, another way to fine-tune the style of generated images. It is based on .pt files that are a few kB in size but can result in massive reproducible visual changes.
This is not textual inversion
This is not dreambooth
This might be similar to hypernetworks ??? I honestly can't tell
Describe the solution you'd like
Implementation, if really new, of the aestetic gradient embeddings use and training
Describe alternatives you've considered
this is new so no alternative besides the other solutions already cited.
Additional context
This is a fork from compvis diffusion too
Is your feature request related to a problem? Please describe. I just discovered the following repository : https://github.com/vicgalle/stable-diffusion-aesthetic-gradients It implements "aestetic embeddings" that are, if I understood well, another way to fine-tune the style of generated images. It is based on .pt files that are a few kB in size but can result in massive reproducible visual changes. This is not textual inversion This is not dreambooth This might be similar to hypernetworks ??? I honestly can't tell
Describe the solution you'd like Implementation, if really new, of the aestetic gradient embeddings use and training
Describe alternatives you've considered this is new so no alternative besides the other solutions already cited.
Additional context This is a fork from compvis diffusion too